Clinical Research Methodology
Aims and Objectives
Be able to understand different types of clinical research/epidemiology methodology
Understand the advantages and disadvantages of different clinical research/epidemiology methods
Use the appropriate clinical research/epidemiology method to investigate research questions
Clinical Research/Epidemiology
Normality/abnormalDiagnosisFrequencyRiskPrognosisTreatment
Steps in Clinical Research
ObservationAssociationCausationInterventionEvaluation
Cross-sectional study
Descriptive study or surveymeasure exposure and outcome in one
moment in timeexposure and outcome are measured
simultaneously
Cross-sectional study
Advantages quick and simple can study many
associations can estimate prevalence low participation/high
response possible to show validity
and reproducibility healthcare planning
Disadvantages problems with casualty survivor bias recall bias inefficient for rare diseases not suitable for disease of
short duration prevalence affected by low
response and migration in and out of population
Steps in Clinical Research
ObservationAssociationCausationInterventionEvaluation
Association and Causation
Association Causation eg earlobe crease and ischemic heart disease
Case control studyCohort study
Explanations for a Positive Association
BiasConfoundingChanceReverse CausationCausation
Bias
A systemic error introduced into the study by an investigator Result from design Just plain wrong
Two main types of bias Selection bias Information bias
Selection Bias
Occurs when selection of cases or control is related to exposure Selection of patients from hospitals, specialised
centres Selection of “healthy” controls from hospitals Response rate bias Self selection bias Survival bias
Selection Bias - example
Large scale study showed that “married” better survival than “widowed”
But … if widowed who remarried are reclassified as “married” and if illnesses reduce the chance of remarried, effect may be due to selection bias
Control from hospitals are more likely to have higher risk of smoking and high alcohol intake
Information Bias
Misclassification
Observer Bias
Recall Bias
Information Bias – Observer Bias
Observer know the underlying hypothesis and ask more probing question to those exposed than controls
Remedies Blind the observer Use highly structured interview
Information Bias - Recall Bias
Disease status affect patients’ response Patient with musculoskeletal diseases are more
likely to remember minor trauma
Particular problem with case control studiesRemedies
Find reliable records Use control with other illnesses
Confounding
Confounders confuse an association
A
C
B
Features of Confounding
Causal relationship between confounder and outcome
Confounder associated with outcomeNot simple on chain of causation
Eg depression smoking MI
Remedies to Confounding
Design Match (match case and control for gender and age) Restriction (limit study to certain groups) Randomisation (limit to treatment)
Analysis Stratification Standardisation Statistical modeling
Case Control Study
Retrospective study of previous exposureIdentify “Cases” and “Controls”Assess and compare the “exposure to risk”
in “Cases” and “Controls”eg smoking and lung cancer, HRT and
ischemic heart disease
Case Control Study
Advantages efficient for rare diseases relatively cheap and quick useful for long latency
periods useful for acute exposure
Disadvantages prone to bias difficulties in selecting
controls inefficient for rare
exposures cannot calculate incidence
rate temporal relationship may
not be clear
Cohort Study
Measures exposure then seek information on subsequent disease experience
PROSECTIVEAvoid bias provide large number are not
lost to follow upBUT don’t remove confounding
Steps in Clinical Research
ObservationAssociationCausationInterventionEvaluation
Intervention and Evaluation
Randomised controlled trial (RCTs)Not all risk can be tested in RCTs
sex smoking income
Clinical Effectiveness Efficacy in trials vs efficacy in real world
Costs, feasibility, acceptibility
Steps in Designing Clinical Trial
RationaleHypothesisType of trialPopulation studiedOutcome MeasuresNumber of casesAnalysis of results
Rationale for Trial
New drug does it work
Established drug are there new indications
Conventional therapy improving efficacy
Delivering care benefits of non-drug therapy
Rationale for Trial
The clinical problemThe burden of diseaseConventional therapy
Benefits Limitations
Rationale for trialStrong Common diseaseHigh health costsHigh social costsPoor current therapySignificant clinical
benefit likely
Weak Rare disease Limited morbidity Reasonable current
therapy Limited benefit
likely
Examples in Rheumatoid Arthritis
Does new analgesic reduce pain?
Does new slow-acting drug reduce inflammation?
Can combination therapy decrease joint damage in early disease?
Hypothesis
Simple
Testable
Relevant
Concise
An Example In RA
Does adding monthly
IM depot steroids to
conventional slow-acting drugs
reduce the number of patients
developing new erosions?
Type of Trial
Parallel Group
Factorial
Cross-over
n- of- 1
Parallel Groups
Cases RandomisedTreatment O
Treatment A
Cases Randomised
Treatment O
Treatment A
Treatment B
Factorial design
Cases RandomisedO A
B A+B
Cross-over Design
Cases Randomised
O A
A O
Selecting Cases (a)
Hospital / community
National/ international
Age, sex and general health
Activity, severity and duration
Select cases (b)Inclusion CriteriaDefinite casesKnown activityKnown durationInformed
Exclusion CriteriaYoung ElderlySickNon-responders
Selecting cases (c)
Hard Inclusion
Criteria
Soft Inclusion
Criteria
Less chance of response
Good generalisability
Rapid entry
Good chance response
Poor generalisability
Long period of entry
Randomised Controlled Trials (RCT)
Randomisation•Tossing a coin•Stratification
sexdisease durationgeographic areas
Blindness•To avoid placebo response•Levels of blindness
Patient blindAssessor blindTriple blind
•Code breaking
Outcome measures
Simple and reliableWidely acceptLikely to changeRelevant for hypothesisOne primary outcomeA limited number of outcomes
Core Data In RADemographic details
Age, sex, disease duration Social class and race
EULAR core data set Joint Counts and Pain Scores Acute Phase Measures (ESR) Function (HAQ) and X-rays (Larsen score)
Predictive factors Rheumatoid factor
Physician And Patient Assessments
Clinical OutcomesImpairment
Persisting synovitis
Functional Limitations in daily living
Radiological Anatomical joint damage
Patient centred Changes in lifestyle
Number of Cases: Sample Size
Calculations are conventional based on showing a difference of 5% significant at 90% power
Continuous discrete variables can be used Additional numbers should be included to
allow for dropouts*
Controls and Blinding
To avoid placebo responsePlacebo or not
placebo no treatmentLevels of blindness
Patient blind Assessor blind Triple blind
Code breaking
Significance and Power
Convention needs a difference with less than 5% of being due to chance
Convention suggests we should be 90% certain that the failure to find no difference between treatments is just due to chance.
Analysis (a)
Always planned in trial design
Usually after all patients completed study
Interim analysis are often misleading
Data dredging is unhelpful
Analysis (b)Potential cases
Did not fulfil entry criteriaGiven information
Did not consent
Randomised
Early withdrawal
Adverse effect
Lack of effect
Inter-current illness
OtherCompleted study
Analysis (c)
Difference must be significant in primary outcome measure for positive trial
Changes in secondary out comes provides supportive information or places therapy in context.